Abstract: Devices and techniques are generally described for fraud detection. In various examples, first data may be received from a remote computing device, the first data specifying at least one of a first internet protocol (IP) address or a first telephone number by a fraud detection service using an application programming interface (API) of the fraud detection service. In some examples, a first machine learning model of the fraud detection service may determine a first confidence score indicating a likelihood that at least one of the first telephone number or the first IP address is associated with fraudulent activity. In some examples, output data may be sent to the first remote computing device via the API, the output data indicating a determination as to whether at least one of the first IP address or the first telephone number is associated with fraudulent activity.

Abstract: Devices and techniques are generally described for fraud detection. In various examples, a first plurality of requests are received over a first time period. In at least some examples, the first plurality of requests may be requests to access a first service. A hierarchical data representation comprising an ordered set of values may be determined for each request of the first plurality of requests. A first subset of the first plurality of requests may be determined. The hierarchical data representation associated with each request of the first subset may include fewer than a threshold number of value substitutions relative to each other hierarchical data representation of the first subset of the first plurality of requests. Access to the first service may be prevented by subsequent requests associated with the first subset of the first plurality of requests.